Emerging
Agentic AI
AI systems that can autonomously plan, use tools, and execute multi-step tasks with minimal human supervision, going beyond simple question-answering.
Explained at five levels
Level 1
AI that can do things on its own — like a robot helper that can look stuff up, use tools, and finish tasks without you telling it every step.
Level 2
AI that doesn't just answer questions but can actually take actions — browsing the web, writing code, sending emails — working like an assistant.
Level 3
AI systems that can autonomously plan, use tools, and execute multi-step tasks with minimal human supervision, going beyond simple question-answering.
Level 4
Autonomous or semi-autonomous AI systems that decompose goals into sub-tasks, invoke external tools and APIs, maintain working memory, and iteratively refine their approach based on environmental feedback.
Level 5
Systems exhibiting goal-directed behavior through recursive planning, tool invocation, and environmental interaction — implemented via ReAct-style reasoning loops, function calling, and hierarchical task decomposition with verifiable sub-goal completion.
Definitions are educational summaries. Terminology can vary by source and context.
Sources
- NIST Trustworthy and Responsible AI Resource Center: Glossary — National Institute of Standards and Technology. Accessed 2026-07-20.